Development and Validation of Models for Preoperative Prediction of Risk and Postoperative Detection of Noninfectious Complications Using Interpretable Machine Learning and Electronic Health Record Data

OBJECTIVE: To apply interpretable machine learning methodology to electronic health record data to develop models for preoperative risk estimation and postoperative detection of noninfectious postoperative complications. SUMMARY BACKGROUND DATA: We previously developed preoperative risk and postoperative detection models for the surveillance of postoperative infections. The purpose of the present study was to develop and validate similar models for the noninfectious complications of the American College of Surgeons National Surgical Quality Improvement Program. METHODS: Preoperative and postoperative electronic health record data from 5 hospitals across 1 health care system (University of Colorado Health), 2013-2019, including diagnoses, procedures, operative variables, patient characteristics, and medications, were obtained. Lasso and the knockoff filter were used to perform controlled variable selection to develop preoperative risk models and postoperative detection models of 30-day noninfectious outcomes of mortality, overall morbidity, bleeding, cardiac, pulmonary, renal, and venous thromboembolism morbidity, nonhome discharge, and unplanned readmission. RESULTS: Among 30,639 patients included, postoperative complication rates for each outcome ranged from 0.1% (stroke) to 10.4% (overall morbidity). The area under the receiver operating characteristic curve for preoperative risk models ranged from 0.68 to 0.91 and from 0.92 to 0.97 for postoperative detection models. Between 6 and 22 predictor variables were included in each model. CONCLUSIONS: We developed parsimonious models for estimating the risk of and the detection of postoperative noninfectious complications. Our models showed good to excellent performance, suggesting that these models could be used to augment manual surveillance.